Class EwmaVolatility

java.lang.Object
com.quantfinlib.volatility.EwmaVolatility
public final class EwmaVolatility extends Object
Exponentially weighted moving average variance (RiskMetrics Technical Document, J.P. Morgan 1996, ch. 5): h_t = λ h_{t-1} + (1-λ) r_{t-1}² on raw (not demeaned) returns. Reacts to volatility regime changes far faster than a rolling window.

Seeding is causal. variances(double[]) has no prior estimate, so for t >= 1 it uses the bias-corrected weights h_t = Σ_{k<t} (1−λ)λ^{t−1−k} r_k² / (1 − λ^t), which use only returns before period t; h_0 has no information at all and is set to h_1 = r_0², the one entry that sees its own period's return. variances(double[], double) takes an explicit prior variance (from a training window) and runs the plain recursion from it. The earlier full-sample seed leaked the whole sample into every early estimate.

  • Constructor Summary

    Constructors
    Constructor
    Description
    EwmaVolatility(double lambda)
     

  • Method Summary

    Modifier and Type
    Method
    Description
    double
    annualizedVol(double[] returns, int periodsPerYear)
     
    double
    latestVol(double[] returns)
    One-step-ahead volatility forecast (per period) for the period after the last return, with the same bias-corrected causal weights as variances(double[]).
    static EwmaVolatility
    riskMetrics()
    The classic RiskMetrics daily decay (λ = 0.94).
    double[]
    variances(double[] returns)
    Conditional variance series aligned with returns: variances[i] is the estimate for period i, formed from information up to i-1.
    double[]
    variances(double[] returns, double seedVariance)
    Conditional variances from an explicit prior: h_0 = seedVariance, then the plain recursion.

    Methods inherited from class Object

    clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

  • Constructor Details

    • EwmaVolatility

      public EwmaVolatility(double lambda)

  • Method Details

    • riskMetrics

      public static EwmaVolatility riskMetrics()
      The classic RiskMetrics daily decay (λ = 0.94).

    • variances

      public double[] variances(double[] returns)
      Conditional variance series aligned with returns: variances[i] is the estimate for period i, formed from information up to i-1.

    • variances

      public double[] variances(double[] returns, double seedVariance)
      Conditional variances from an explicit prior: h_0 = seedVariance, then the plain recursion.
      Parameters:
      seedVariance - variance estimate for period 0 formed before the sample starts, > 0 and finite

    • latestVol

      public double latestVol(double[] returns)
      One-step-ahead volatility forecast (per period) for the period after the last return, with the same bias-corrected causal weights as variances(double[]).

    • annualizedVol

      public double annualizedVol(double[] returns, int periodsPerYear)